Core idea

how do they tackle the curse of dimensionality?

When to use them

When not to use them

Layers

Pytorch implementation